Conventionally, a bundle adjustment technique can be used to construct a three-dimensional (3D) model of an object that appears in one or more images. For example, the images may comprise photos or frames from a video. However, the images input into the bundle adjustment technique may be blurry and may degrade the accuracy of the resulting 3D model of the object.